Study On Fire Intelligent Video Monitoring Recognition Techniques Based On Image Processing

Fire brings huge harm,which involving a wide range,and the reasons of fire become more complicate.Especially the large space building and underground building's fire monitoring and warning,the traditional fire detector can not put into its performance effectively.Because large space building and underground building have large space and many obstacles,fire smoke is difficult to reach the top of a house when fire happens.As a result,it can not be discovered in the early time and give alarm to us in time,we are not able to find out the fire position usually.Thus linkage device become useless.It's too late to alarm when fire become to be out of controll,the great loss is caused.Aiming at traditional fire detector defects,fire physical characteristics and its special performance is mainly studied in this thesis when fire happens.Because the fire images have good property of real.time,and smoke has lots of fire feature information,we proposed fire intelligent video monitoring recognition techniques based on image processing by using image processing technology and wavelet analysis theory.The method firstly gets fire images from video monitoring system and converses it from RGB mode to Gray mode,then the fire images are processed by some preprocessing methods such as image enhance,image smooth and sharpening,thirdly the early fire image features are extreated to make recognition whether fire happens or not.If fire happens,fire linkage device will be started to alarm by the intelligent system decision.In viture of the smoke image,two approaches are put forward in this thesis:image fire detection method using the wavelet transform(IFDMWT) and image fire detection method using frame.image subtraction(IFDMFS).In IFDMWT,the abtained video images are proposed firstly to improve the image quality,then the images are de.composed with wavelet transform,the horizon,vertical and diagonal component in high frenqency band are used as feature vector.The decision are taken based on the feature vector whether fire burst or not by use of ANN.In another method,IFDMFS,image subtract is applied.In order to get the fire features,the image acquired under the condition without fire occurrence is use to be substracted from video monitoring image,the difference is regard as the feature vector which is use to detect where fire is appearance or not.In these two methods,the fire detection is treated as the problem of 2.class classification by use ANN.One is the condition without fire,another is the condition of fire existence.The video monitering experiments prove that the two methods are accurate, have good property of real.time and purness.They also have important scientific theory purpose and application value of engineering.